Early Inquiry of Interplay Between Business Cycles

It is a long debate on whether business cycles have empirical regularities. With respect to these empirical regularities, there is a significant difference between the view of modern business cycle researchers and that of their classical predecessors. In the classical tradition, the business cycle is an endogenous mechanism where fluctuations were seen as a recurrent phenomenon with characteristic periodicities. An important

economic series. This tradition originated in the 19th century with the work of Juglar (1862) and has continued until the 1950s, inspiring among many others, the works of Kitchin (1923), Kuznets (1930), Kondratieff (1926) and Schumpeter (1939).

In contrast, the perception of most modern macroeconomists, whom were deeply influenced by Burns and Mitchell (1946), is that the mechanism of business cycle is exogenous, where economic time series typically do not have a pronounced regular cyclical pattern. According to the modern view, the only defining property of business cycles is the strong coherence of many important economic time series, i.e., their tendency to move together (Sargent, 1987). Therefore, research about modern views, eagers to find out the determinant of the relationships of different economic variables, lead to theoretical explanations such as literature of the real business cycle (RBC). They believed the business cycle is a random process; only exogenous shocks can generate the business cycle phenomenon. For example, Kehoe and Prescott (2002) use the RBC framework to explain the episodic ‘‘Great depressions’’.

However, even the modern view has offered a more comprehensive framework than the traditional view to analyze the interplay of different economic variables over the business cycle, it is still insufficient. Since the general perception is that recessions exist because there were expansions before it, therefore, boom and bust should be considered together with a holistic framework. Though the modern view can explain the cause and consequence of the recession, it is incapable to holistically consider the boom and burst together.

On the other extreme, famous researchers have established plenty explanation for the recurrent boom and bust through the traditional view－periodic and coexistence of different types of cycles. The theoretical driving forces of different kinds of cycles are

regular fluctuations of investment activities: the Kitchin with inventory investment, the Juglar with investment in machinery and equipment, the Kuznets with building or transportation investment and then the Kondratieff with the construction of basic capital goods which is lead by clusters of innovations such as railways and canals in the sense of investment. The different cycle lengths are each associated with the particular form of investment, determined by the durability of the investment and the time lags between movements in final demand and the completion of the invested capital good (Duijn, 1983).

Noteworthy, considerable debates about such regularity of business cycles have always been there, since their lengthy durations always accompany institutional and economic structural change (Abramovitz, 1968; Solomou, 1998; Maddison, 1991). Thus, past experience might not yield value to the future. Particularity, Abramovitz (1968) had argued that the Kuznets cycle has disappeared since the Trans-Atlantic migration ended. Besides, insufficiency of data is another reason against the traditional view.

Burns and Mitchell had noted in their comprehensive work Measuring Business Cycles (1946) that since their own data only extend to the 1930s, which is too brief a span to determine whether building cycles (Kuznets cycles) and Kondratieff cycles were a continuing feature of the modern economy. Becker in his presidential address of American Economic Association (AEA) in 1987 echoed Burns and Mitchell’s consideration. “If long cycles of the Kondratieff or Kuznets type exist, we will need another 200 years of data to determine whether they do exist or are just a statistical figment of an overactive imagination”.

However, if we look into the theoretical cause, most the mechanisms of business cycles of different frequencies still validate. Even if the amplitudes and frequencies of

these different cycles are not stable due to varying economic environment and structure;

their forces still work in modern economies. Therefore, the length of the cycle and also the amplitude to some extent are variable, however, their variations taking place within limits (Frisch, 1933; Hillinger, 1992). Especially to answer doubts if the Kuznets cycle had vanished, Easterlin (1987) asserted that even the Trans-Atlantic migration has waned, interplay of inter generations may still make the Kuznets cycles vivid.

Past research also provides plenty of evidence to support the traditional view. It was Kondratieff (1926) who first conceived the coexistence of shorter- and longer-term cycles and the corresponding effect of their interplays. He argued that during the rise of the long waves, years of prosperity are more numerous, whereas during the downswing, years of depression predominate. Dujim (1985) used the industrial production data of UK, US, West Germany, France, and Japan and Shinohara (1996) used post war Japanese data to confirm such concept. In addition, Schumpeter (1936) not only echoed the coexistence of short and long cycles, but formally distinguished them as Kitchin cycles (3~5 years), Juglar cycles (7~11 years), Kuznets cycles (15~25 years) and Kondratieff cycles (40~60 years). In this respect, Reiter and Woitek (1999) used data of 15 OECD countries and found that a large number of their covered countries experienced regular Kitchin and Juglar cycles in the period 1960~1993. Berry (1991) used real and nominal series data of the US and UK and also validated the regularity of Kuznets and Kondratieff cycles.

Furthermore, earlier research also used the idea of coexisting cycles to explain the deep recessions in economic history. Schumpeter (1936) pointed out that deep recessions in the period covered by his material, namely 1825~1830, 1873~1878 and 1929~1934, all came in times when all cycles were in their downward phases. Similar

arguments of coincidence in the downturn of two or more cycles (among Kitchin, Juglar, Kuznets and Kondratieff cycles) have also been suggested by Berry (1991) on the Great Depression of US and Sen (1997) on the 1990s Russia and Shinohara (1996) on the 1990s Japan.

Therefore, even if the interplay between business cycles is not fully agreed among the research community, however, in the description of economic evolution, it is inadequate to dismiss the force of different types of cycles in understanding the process of the economy. Besides, we have 70 more years of data than what was available to Burns and Mitchell; therefore we are in a better position to discuss the regularity of cycles in the traditional view than they once were.

4.3. Data

The data we use in this section comes from two sources; one is the International Financial Statistics (IFS) database, where we take the quarterly industrial production data of 15 OECD countries and aggregate advanced economies from first quarter 1961 thru first quarter 2009. The 15 nations are: Austria, Belgium, Denmark, Finland, France, Germany, Italy, Japan, the Netherlands, Norway, Spain, Sweden, Switzerland, UK and the USA. The other source is the real GDP data of the aforementioned countries and their aggregate from 1870 thru 2006 edited by Madisson (2009), where we merged it with IFS data to obtain the real GDP data of major countries and their aggregate from 1870 thru 2008.

The reason why we use two types of data to examine the interplay of different cycles is due to several considerations. First, the cycles discussed in this section include

the 3~5 year Kitchin, 7~11 year Juglar and 15~25 year Kuznets, though high frequency quarterly data contain more information, as the duration of the cycle type being verified increases, the time length of the data set might not satisfy the basic requirements of spectral analysis. For instance, the industrial production data we use is only 48 years plus 1 quarter in length, which may span over eight to fourteen 3~5 year Kitchin cycles, and four to six 7~11 year Juglar cycles, but it only covers two 15~25 year Kuznets cycles and one Kondratieff cycle at best. However, some scholars consider the length of data necessary for applying spectral analysis must at least cover three cycles (Klotz and Neal, 1974), some others thought seven cycles were the minimum requirement (Granger and Hatanaka, 1964), while still a few suggest at least ten cycles (Soper, 1975) worth of data were needed to perform spectral analysis

Thus, even with the more relaxed demands of Klotz and Neal (1974), quarterly industrial production data is not long enough to discuss the 15~25 year Kuznets cycle and the 40~60 year Kondratieff Cycle. But by merging the Maddison (2009) and IMF data, we can obtain a 139 year series for real GDP, which covers seven to nine Kuznets cycles, and satisfies both Klotz and Neal (1974) and Granger and Hatanaka’s (1964) requirements. However, even with annual GDP, the data length is still incapable of analyzing the 40~60 year Kondratief cycle, the reason why in verifying the existence of cycles in this section, we will mainly focus on the Kitchin, Juglar and Kuznets cycles.